Advancing AI: Highlights from January

At Sony AI, every month is an opportunity to innovate, connect, and share our progress with you. Welcome to Advancing AI, our brand-new monthly round-up series where we highlight the research, events, and people shaping the future of AI.

Let’s dive into the highlights from January, and set the stage for what’s coming next.

Latest Research

Three key innovations were presented by Sony AI teams at NeurIPS 2024 in December:

LOCKEY: Securing Generative Models
Generative models are increasingly susceptible to misuse. The LOCKEY system integrates key-based authentication, embedding user-specific watermarks for secure and traceable outputs while preserving audio quality. This approach establishes a new benchmark for securing generative models.
👉Read the full paper

DUPLEX: Reinforcement Learning for Diverse Behaviors DUPLEX introduces diversity-focused reinforcement learning methods, enabling agents to discover multiple near-optimal strategies for dynamic environments. Demonstrated in Gran Turismo™ 7 and robotics tasks, it enhances adaptability and performance in real-world applications.

👉Learn more about DUPLEX

DECO-Bench: Advancing Privacy in AI Addressing the critical need for privacy-compliant AI, DECO-Bench establishes a unified framework to evaluate privacy-preserving data methods. It integrates synthetic data generation and structured data splits to balance privacy and utility.
👉Explore DECO-Bench

Sony AI Global Events

At NeurIPS 2024, Sony AI showcased a diverse portfolio of research spanning ethical AI, generative models, reinforcement learning, and privacy-preserving machine learning. The conference provided a platform for discussing challenges like dataset fairness and the evolving landscape of AI development. Highlights included a focus on practical applications such as secure generative AI (LOCKEY) and advanced music synthesis tools (DisMix).
👉 Read on: From Data Fairness to 3D Image Generation: Sony AI at NeurIPS 2024

Researcher Spotlight

Patrick MacAlpine and GT Sophy’s Evolution

Patrick MacAlpine has been a research scientist at Sony AI since August 2020, contributing to the Gaming and Interactive Agents Flagship. Leveraging his expertise in autonomous multiagent systems, robotics, and machine learning – with an emphasis on reinforcement learning – MacAlpine has focused on advancing research on Gran Turismo Sophy (GT Sophy), the groundbreaking AI racing agent a unique collaboration between Sony AI, Polyphony Digital, the creative studio behind the world-famous Gran Turismo games, and Sony Interactive Entertainment. MacAlpine and the Gaming and Interactive Agents team recently showcased their latest research at the 2024 Gran Turismo World Series World Finals last month in Amsterdam, where human drivers tested GT Sophy’s exploratory capabilities, including its ability to manage tire wear and fuel consumption.

MacAlpine, along with Kaushik Subramanian, recently discussed the new research and capabilities of GT Sophy with GTPlanet. 👉Watch here

Celebrating Alice Xiang’s Prestigious Privacy Award

We are thrilled to announce that Alice Xiang, Lead Research Scientist at Sony AI and Global Head of AI Ethics at Sony Group Corporation, has been honored with the Future of Privacy Forum’s Privacy Papers for Policymakers Award for her groundbreaking paper, Mirror, Mirror, on the Wall, Who’s the Fairest of Them All?. This award recognizes her exceptional work in addressing systemic bias in AI systems and her thought-provoking analysis of fairness as a reflection of societal patterns.

Her research will be featured in the Privacy Papers for Policymakers Digest and presented at the 15th Annual Privacy Papers for Policymakers event in Washington, D.C., this March. This achievement underscores Alice’s dedication to fostering ethical, fair, and accountable AI.
Learn more about the award.

Stories from the Sony AI Blog

Sony AI Ethics Flagship: Reflecting on Our Progress and Purpose
In January, Sony AI celebrated the milestones of its AI Ethics Flagship. Under the leadership of Alice Xiang, this initiative has advanced research in bias mitigation, fairness evaluation, and the interplay between privacy and fairness. Highlights include the TAB framework for addressing dataset bias and the development of inclusive evaluation metrics.
👉 Read on: Sony AI Ethics Flagship: Reflecting on Our Progress and Purpose

Did you miss it? Check out our most popular blog from December: Exploring the Challenges of Fair Dataset Curation
Sony AI's recent paper, "A Taxonomy of Challenges to Curating Fair Datasets," presented at NeurIPS 2024, dives into the complexities of building fair and inclusive machine learning datasets. The research emphasizes fairness not as an afterthought but as a principle integrated throughout dataset curation. By addressing real-world implications, such as biases in healthcare and hiring systems, the study proposes systemic changes that prioritize equitable outcomes.
👉 Read on: Exploring the Challenges of Fair Dataset Curation: Insights from NeurIPS 2024 – Sony AI

Sony AI in the News

AIHub: Yuki Mitsufuji sat down with Dr. Lucy Smith, Managing Editor of AIHub.org, to discuss his team’s most recent research into image generation presented at the 2024 Conference on Neural Information Processing Systems (NeurIPS): GenWarp and PaGoDA. 👉Read here

The Road to Accountable AI: Alice Xiang spoke with Kevin Werbach, Host, Professor, and Chair of Legal Studies and Business Ethics at the Wharton School, University of Pennsylvania, about the current landscape of AI ethics, challenges organizations encounter regarding algorithmic fairness, and potential barriers that could be encountered in AI ethics and governance.
👉Listen here

Traxion: Kaushik Subramanian and Patrick MacAlpine caught up with Thomas Harrison-Lord after the Gran Turismo World Series 2024 World Finals to discuss the newest exploratory capabilities of Gran Turismo Sophy (“GT Sophy”) and the agent’s strategy.
👉Read here

What's Next for Sony AI

We’re proud to showcase two accepted papers at the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI-25), held from February 25 to March 4, 2025, at the Pennsylvania Convention Center in Philadelphia. As one of the leading global forums for artificial intelligence, AAAI-25 will host cutting-edge technical presentations, workshops, and special tracks spanning the breadth of AI research and applications.

Coming up in March, we will be back at SXSW to present two panels.

Details:
SXSW — Pushing Creativity to New Bounds: Future Robot Applications

📅 Date: March 10, 2025
🕓 Time: 4:00 PM – 5:00 PM CT
📍 Location: Hilton Austin Downtown, Salon K

Join an insightful panel discussion exploring the intersection of robotics and creativity. Peter Stone from Sony AI, alongside other experts, will delve into how advanced robotic technology is reshaping creative disciplines like music, art, and comedy. The session will also highlight the importance of fostering responsible development through collaboration with creators. A must-attend for anyone interested in innovation and the future of creative technology.

SXSW — Can Human & AI Collaboration Extend the Bounds of Creativity

📅 Date: March 12, 2025
🕙 Time: 10:00 AM – 11:00 AM CT
📍 Location: Austin Convention Center, Room 9ABC

This fireside chat will examine how AI is pushing the boundaries of creativity while maintaining respect for traditional mediums. Michael Spranger from Sony AI and Dawn Chmielewski from Reuters will explore the balance between technology and artistry, showcasing the potential for AI to redefine creative possibilities. Perfect for creators curious about the evolving collaboration between humans and AI.

Join the Conversation

We’d love to hear your thoughts on our January highlights and the new monthly round-up series! Connect with us on LinkedIn, Instagram, or Twitter, and let us know what you’d like to see in future editions. Until next month, keep imagining the possibilities with Sony AI.